Study design
This study was a multicenter, cross-sectional pilot study of 250 women and their 257 infants, carried out at the Korle-Bu Teaching Hospital (KBTH) in Accra and the Tamale Teaching Hospital (TTH) in Tamale, Ghana. All eligible participants for the study were approached and recruited if they presented at the 1 or 2-week postnatal clinics. Mothers and infants were excluded from enrollment if the infant was younger than 3 days or older than 30 days old. Additionally, mothers with known chronic diseases that could impact urine iodine concentrations, such as chronic kidney disease or history of intake of iodine-containing medications, were excluded.
Definitions
A newborn screen for congenital hypothyroidism was considered positive if the infant filter paper TSH was > 17 μIU/ml in whole blood [20, 21]. Based on the World Health Organization’s (WHO) classification, maternal urine iodine concentration was further disaggregated into: severe iodine deficiency (< 20 μg/L); moderate iodine deficiency (20–49 μg/L); mild iodine deficiency (50–99 μg/L); sufficient iodine level (100–199 μg/L); above requirements (200–299 μg/L); and excessive iodine intake (> 300 μg/L). Mothers were considered iodine deficient if their urine iodine concentration was < 100 μg/L [16].
Recruitment/procedures
In Ghana midwife-obstetrician teams oversee the first two post-natal visits, typically at 1- or 2-weeks post-partum and again at 6-weeks post-partum. At the beginning of each clinic, midwives lead a group class on a topic of their choice. Research nurses for this study, midwives who work at their respective sites, led the group class during recruitment periods. They taught clinic attendees about the potential effects of maternal iodine deficiency on children and introduced the research study during this group class. After the class, women who indicated interest in the study were further screened to determine eligibility. Mothers who came to the clinic after the class was over were also approached individually, informed about the study and screened for eligibility, if they indicated interest.
Informed consent was obtained from mothers in English or in the local languages of Ga, Twi, Dagbani, and Hausa, as needed. Urine samples were collected from mothers, transferred into small tubes and frozen within 6 h of collection. Frozen samples from each site were then batched and sent to the urine iodine lab at the end of the 8-week data collection period. Samples from both sites remained frozen throughout the shipping process.
The urine iodine concentration test was run by the Micronutrient Research Laboratory of the Department of Nutrition and Food Science of the University of Ghana. This lab is certified by the U.S. Centers for Disease Control and Prevention Ensuring the Quality of Urinary Iodine (CDC EQUIP) program for iodine laboratories, supporting numerous projects and countries. Urine iodine concentration was determined using the Sandell-Kolthoff reaction [17].
Heel-stick blood samples from infants were collected on filter paper. The filter paper cards were labelled, dried for at least 1 h, and placed in individual Ziploc® bags with desiccant. Samples in both locations were stored in a cool, dry, secure location. In Accra, the samples were sent directly to the laboratory within 6 h of collection. All samples from Tamale were stored in a laboratory within 6 h of collection, then batched and shipped to Accra at the end of the data collection period. TSH levels were determined using ELISA assays on the dried blood spots. The TSH ELISA kits were manufactured by RayBiotech, Norcross, GA. The kits were shipped to Ghana on dry ice via commercial courier service, and the assays were run by the Chemical Pathology laboratory at the Department of Medical Biochemistry, School of Basic and Allied Health Sciences.
We collected data from mothers, including demographics such as maternal age, age of infant, gestational age of infant and history of multiple gestations; nutritional information such as whether iodized salt is used in the home; and clinical information, for example, medications taken in pregnancy, complications of pregnancy and delivery, history of goiter, and delivery type [22]. Research staff reviewed maternal antenatal and delivery charts to obtain information to supplement the participant interview. Data were managed using the Research Electronic Data Capture (REDCap), an electronic data management system hosted at Vanderbilt University [23].
Power analysis
We used the Pearson correlation coefficient method for power analysis. A sample size of 150 provides at least 90% power to detect a clinically significant correlation between maternal urine iodine levels and infant TSH, with a correlation coefficient of 0.6 and a 2-sided type 1 error of 5%. Hence, a sample size of 250 provided more than adequate power to detect a difference between the two populations. We estimated the mean and standard deviation of TSH levels in Tamale and Accra will be approximately 9.82 ± 1.64 and 4.18 ± 1.17, respectively [24]. Our secondary aim was to compare congenital hypothyroidism prevalence between infants in Accra and Tamale using TSH values as a measure of congenital hypothyroidism.
Statistical analysis
Descriptive analyses were conducted on demographic and nutrition data. Categorical variables were described using proportions and most continuous measures were described using means and interquartile ranges (IQR). Maternal urine iodine concentrations and TSH levels were described using medians with 95% confidence intervals (CI) calculated by bootstrapping, as per the United Nations Children’s Fund (UNICEF) recommendations [25]. We used Chi-squared, t-tests or Wilcoxon rank tests as appropriate for comparisons. A Spearman correlation coefficient was employed to evaluate the correlation between maternal iodine levels and infant TSH.
We then assesed variables which we hypothesized would be associated with infant TSH using univariate linear regressions models. These variables included: maternal urine iodine concentration (log-transformed to account for its skewedness), maternal age, location (Accra or Tamale), mother’s educational status (dichotomized into none or some formal education, with some formal education defined as any amount of schooling from lower primary school (class one to class three) through to tertiary education), parity (primiparous/first-time or multiparous), infant age (in weeks), infant birthweight, infant sex, maternal use of iodized salt, maternal use of bouillon cube and maternal seafood intake. Variables significant at the P < 0.1 level in the univariate analysis were included in a multivariable regression model to identify the factors with independent effect on infant TSH. P < 0.05 was considered significant for the multivariable regression.
Further, we performed sensitivity analysis by excluding outliers of maternal urine iodine concentration and by dichotomizing maternal urine iodine concentration into high and low iodine levels, using the WHO cutoff of 100 μg/L. We also attempted to conduct sensitivity analysis of infant TSH by dichotomizing TSH into < 10 μIU/ml and ≥ 10 μIU/ml. We used TSH of 10 μIU/ml for this analysis because normal infant TSH in the first month of life is < 10 μIU/ml across many different populations [3]. However, only 13 out of 258 infants for whom TSH data was available fell above the cutoff of 10, thus further regression modeling on this small sample was not performed.